Bitcoin Forum

https://i.imgur.com/EmpXi.jpg

I was sitting waiting patiently, a whole hour, for a new block to be found and watching this chart when it occurred to me that the irregular creation of blocks will eventually become much more regular.

Right now, miners are motivated by the 50 bitcoin reward and hardly at all by the fees, although those are nice too. In the future when there is little reward and high fee income, it will not be worth the electricity cost to mine shortly after a block is found because the number of transactions will be low and will take time to accumulate. After several minutes (less than ten) it would become worth the cost to mine for the fees. The result is that almost all miners individually, acting rationally, would stop mining after a block is found and simply go about routing and keeping track of transactions until enough fees accumulate to warrant the electricity to search for a block. Then they would all start again.

So after a block is found, mining would stop for several minutes and then quickly start back up again 3 to 6 minutes later.¹ The difficulty would take this behavior into account so that finding a block after those couple minutes will be easier and the ten minute average would remain. No more would we get 2 blocks right next to each other² which would mean fewer times, like pictured above, when an hour goes by without a confirmation.

And the best part is that we don't have to do anything to make it happen! Yay!


1: 3 to 6 minutes is a guess but it would definitely be less than ten.
2: And if there is, it would only be because other miners weren't aware of the new block which means that the block chain is split. One of the two blocks will be orphaned which means that it effectively was never found in the first place.

There are plenty of miners out there that don't pay for power and hence won't care about whether the block is full or not.

That's beyond the fact that I don't see transaction fees funding things very well at all, unless the network forces higher fees or the price of bitcoin goes up a lot.

I have a book for you to read. Fooled by Randomness. It is aout trading ,but it is about how people have no clue about statistics.

> So after a block is found, mining would stop for several minutes and then quickly start back up again 3 to 6 minutes later.

Yes. Like thousands of miners are on the computer turnin thigns off aftera block is found.

Here is what goes on:

* It takes time for a block and work to go through the network, and miners to pick up new work. not everyone uses long poll, but even then there is a lot of communication - so once a block is found, a lot of miners still work on outdated data and get no shares.
* Block creation will never stabilize and  be predictable because it is RANDOM. One block per 6 minutes is AVERAGE. The variance is quite high.
* And yes, you have fluctuations of hashing power, but it is not because a block is found, more because of weather and people tuning off stuff at certain times, people not able to sleep with a mining rig running in their room, but mining during the day. THIS will no change, although at one point other people may be more dominant running data centers.

I think it would be easier to lower the target time to like 5 or 3 minutes.

The effect will exist, but its strength will be less than you describe because:

1. Depending on the mechanisms that will be put in place to ensure inclusion in a block is a scarce resource, even after a block is found there will still be plenty of fees in the next.
2. Unless there's something else to compute, the decrease in profitability will have to be significant to pause mining, to avoid thermal cycles.
3. Some people just won't care enough about this to pause mining.
4. Even if the numbers are as you describe, instead of exp(10) we'll have 5+exp(5), which still leaves plenty of room for longer than average blocks.


They had better. But it's possible that in the future proof-of-work will be augmented with things like proof-of-stake so not a lot of mining will be necessary.

Yes.

We're hoping Bitcoin will succeed, which will inevitably cause its price to increase.


These days there are things called "computers" which can automatically halt the mining when a new block is found.

Propagating a block is supposed to take a minute tops. And this is completely irrelevant to the OP's point - people will mine on the old block, but if they find something their block will be invalid. Thus the finding of valid blocks will slow down.

He didn't say it will be completely predictable, only that it will be more predictable because it will no longer be a homogenous Poisson process.

Did you read the post? He explained exactly why this will change.

They obviously won't do it manually. Their mining software will do it.

Nothing in your post refutes what I have said.

That's true. They will happen. I am happier about many fewer occasions.


This statement makes this kind of assumption for example: 'It is worth it to mine after one minute of the previous block being found.' If this assumption is true, then it will be twice as 'worth it' after an additional minute. After ten minutes, finding a block would then be worth ten times as much as the effort put in. This is not a stable situation- many more people would start mining and the benefit of mining would drop until your assumption becomes not true.

This is great. I got chills back when I first realized. I think we'll start to notice it to a small degree in 10 years when 6.25 is the base reward. The free electricity people won't stop, but a good number of people will and we'll be able to see it in slightly more even times between blocks.

Actually, it probably depends on what it does to hardware to change the load on it constantly. If that's bad for hardware then we probably won't notice it for longer.

Secondary uses for hashing power will influence things too. Even if you have free electricity you might switch between bitcoin and namecoin based on the accumulated fees waiting on each.

There is no "acting rationally" in social behavior driven by individual greed of each participant.  You may stop, guy next to you will be in the race trying to get ahead of you.

By "acting rationally" I did indeed mean acting selfish and greedy and making rational decisions to increase your own profits.

block creation is perfectly stable and predictable.  it was designed that way.

but it was designed by someone with a long-term vision, and you simply can't look at it in the short term.

Um, it is not perfectly stable and anyone reasonable looks both short and long term. Are you missing the point of the original post?

it simply will not happen how you describe it.  we don't have to go far in history for examples. take current political and economic systems in the world. are our countries (governments) acting rationally and in the best interests of their citizens or hugely driven by special interest lobbying groups looking after their corporate sponsors? is news media presenting real news or rationalizing and dramatizing  everything to point that it isn't objective news anymore but a show to pump ratings?  I tend to think we live in the age of idiocracy especially here in the US  and the reason for all of this is human greed. I seen whole industry destroying itself driven by individual greed, if any one wonders, I'm talking about online porn industry. There is no acting rational in mass driven by greed and profits, can't see it working like that.

you mean this:


?

no.  i don't think i'm missing it - i just don't think it's valid.

on a practical level, no miner does this.  it's a ridiculous amount of effort.

and on a strictly financial level, i suspect it's more expensive (in terms of both electrical power and wear on equipment [starting and stopping fans, etc.]) to do this.  do you turn off the engine of your car, waiting at a stoplight (some new hybrids - which are designed for that - excepted)?  no.  you use more gas starting an engine than you use in 30-60 seconds of idle.

i just can't see starting and stopping my miners six times an hour as a particularly efficient strategy.  even if i were to write a script to automate it, it would still - i offer - use more power and degrade the equipment more than would be offset by any tiny, theoretical profit.

as long as you don't have more than a dozen, it would probably not waste much electricity at all.a fan does not have much mass so starting it would would not be too much burden.

Google "Poisson distribution". Get educated.

it's not just the fans on the GPUs.

put a Kill-A-Watt on a miner sometime, and compare the power draw in the first five minutes of operation to the draw at steady-state.

write up a lab report for me and ill belive you. you fail to state the state the "miner" was in before you started the meter. stopping and starting a mining operation, as long as all the data stays in ram, and the gpu just stops doing numbers, hardly any energy would be wasted by not mining, other than keeping all the equipment idle. with 3-5 gpus the savings would be great because you only have 1 set of ram, cpu and mobo and 3-5 gpus that won't be working. with only 1 or 2 gpus on a mobo i could see it as being a waste, but with 3-5+ i could see it working assuming the original assumption was correct.

How about in 140 years? Do you think people will mine if they get 0BTC for finding a block? Or maybe they'll wait until there is a fee offered causing the effect the Atheros is talking about.

Another cool effect is that you'll actually be able to get faster service with a higher fee even once the fee you were going to pay guarantees inclusion. If you pay .001 you'll get in for sure, but if you pay .1 maybe you'll make the difference for 10000 miners and decrease your expected wait by 30 seconds. There probably will be some roughly known point at which all available hashing power is on and extra won't help or at least give greatly diminishing returns. So watching the basket of fees will be part of excellent fee strategy.

i have two GPUs/mobo.  the reason for that is that it's a much more cost effective build.  the PSU to run a couple of 5870s doesn't scale at twice the dollars to running four - that PSU is much more than twice as expensive.  ditto for a two PCIe-slot mobo vs. a four (really five) slot mobo.  and etc.

hardware cost doesn't scale arithmetically - it scales geometrically.

so yes, i can see it working for a 3-5 GPU setup too.  but the upfront costs would be ridiculous.  talking about the comparatively minuscule differences in profitability due to starting and stopping vs. not starting and stopping would be a waste of time - if one took all costs into consideration.

There is something that may mitigate this somewhat: Most power plants can't start and stop instantaneously. Nuclear power is the slowest and often only used for baseload for that reason. Industrial miners using solar/wind power would face storage or transmission losses if they elect to wait before trying to process a block.

I think even the fastest power plants have a turn-around time of at least a minute.

Read the OP. Then make an on-topic comment.

Block finding is a non-homogeneous Poisson process (http://en.wikipedia.org/wiki/Non-homogeneous_Poisson_process) with rate parameter proportional to the worldwide hashing power dedicated to finding a valid block. If it was homogeneous, time between blocks would follow the exponential distribution. The OP is suggesting that, due to the financial incentives of the miners in a post-minting world, the hashrate will drop after every block and gradually climb back, making the time between blocks distribution less variable than the exponential.

Thank you for your replies, especially FreeMoney and Meni Rosenfeld.

So it seems definite that this will happen to large degree provided that the thermal cycles are not damaging to the hardware. If a secondary equally rewarding use of computing power exists then this behavior, which I will call "miner-pause behavior" will definitely happen. Block generation will slowly switch from being a homogenous Poisson process to a non-homogenous Poisson process. We should be able to calculate the exact amount of time after a block is found that it is worth it for miners to start mining but the fact that people pay different amounts for electricity, including 0, would make it quite complicated to estimate at this time.


I accidentally worded that very officially. :-|

how would pool hopping play into this,- jumping after the first 43.5%...

I'm not completely familiar with pools but I've just read that "Proportional pools try to combat [pool hopping] by not immediately publishing when a new round has begun but this is not 100% effective. " In this pool, you would want to mine all the time. However all of the miners would be at a disadvantage in this pool because no one could do 'miner-pause behavior'. Individuals would be smart to abandon this kind of pool and do solo mining or a type of pooled mining where miners can all benefit from miner-pause behavior. If no one comes up with a fair way to do pools AND miner-pause behavior, then large miners would never want to be in a pool.
But I don't know much about pools yet. I don't know how they keep track of the work miners do.

That's right. The answer is to use the current power method instead of the persistent shares. The very first pool did this, but that didn't win out for psychological reasons. People hate not getting paid because they disconnected for a minute, but it doesn't really matter since it evens out so fast. Serious profit maximizing miners will move to current power pools as soon as this effect starts to show.